Why Are Your Employees Hiding Their AI Use?

Because admitting it costs them. Controlled research on knowledge workers found that people who disclose using AI on a task get rated dramatically lazier than peers who turned in identical work, and meaningfully less likely to be recommended for high-visibility projects. Same work, one difference: the evaluator was told AI helped. The problem isn't adoption. It's admission. We call the price of admitting it the disclosure tax, and it's the most expensive tax in your company right now, because it falls hardest on exactly the people spreading AI capability through your organization, and it convinces them to stop.

What is the disclosure tax?

It's the reputational price an employee pays for saying AI helped with their work.

The evidence here is unusually clean. In a controlled experiment, evaluators reviewed identical pieces of work, with one variable: whether they were told AI was involved. When it was, they judged the creator as far lazier and were substantially less likely to put them forward for important projects. Follow-up research found the penalty holds across occupations, ages, and genders. Now put that next to the adoption numbers, where the overwhelming majority of knowledge workers use AI on the job, and the picture resolves into something strange: nearly everyone is doing the thing, and everyone is still punishing the people who say so. Your employees have run this math. Keep the productivity, skip the mention. The work gets faster and the company learns nothing about how.

Why does the tax cost the company more than the employee?

Because absorption spreads through demonstration, and the tax makes demonstration irrational.

AI capability doesn't diffuse through a company by memo. It spreads when one person figures out how to make a tool work for your specific context, shows a colleague, who adapts it and shows the next person. That chain is the entire mechanism behind closing the adoption-to-absorption gap, and every link in it requires someone to say, out loud, here's how I actually did this. The disclosure tax severs the chain at the first link. Your most capable AI users keep compounding their own advantage in private while the rest of the organization stays exactly where it was, and leadership reads the flat results as proof the technology underdelivers. It doesn't stop there. People who won't disclose also won't surface what tools they need, which is how you end up with shadow use: employees running work through unsanctioned tools rather than admitting the sanctioned ones fall short. The tax leaves AI use fully intact and takes away your visibility into it instead.

Who actually levies the tax?

The people who don't use AI themselves. That detail changes everything about the fix.

The research found the laziness penalty comes from evaluators who aren't AI users, and it fades when the person judging uses AI in their own work. Read that as an org chart: the tax is levied by the lagging half of your company on the leading half. Every skeptical manager who's never opened a model becomes a toll booth, and ambitious employees route around it the rational way, by going dark. The cruel part is the direction of the judgment: the people best positioned to teach the organization are being graded by the people with the least basis to grade them, and the grade sticks to careers. This is a close cousin of the manager bottleneck, where adoption stalls in the middle layer regardless of what the top funds and the bottom wants. The difference is that the bottleneck slows approval, while the tax poisons something more basic: the willingness to be seen. You can mandate approval. You can't mandate what people admit.

Can policy fix it on its own?

No. Policy makes AI use legal. Only culture makes it visible.

Plenty of companies have responded to shadow use with clearer rules, sanctioned tool lists, mandatory training, and acceptable-use policies, and those things matter, but none of them touch the tax, because the tax was never levied by the policy. It's levied in performance reviews, in who gets tapped for the visible project, in the joke that lands in standup, and in the flicker of judgment when someone says a model drafted the first pass. Here's the finding leaders should sit with: in organizations where leadership actively and publicly celebrates AI use, the penalty nearly disappears, and disclosed AI users start getting rated as more capable than the people who stay quiet. The tax rate, in other words, is set by leadership behavior: what gets praised in public, who gets tapped for visible work, whether executives use the tools openly, and how the room reacts when someone says a model helped.

What should leaders do about it?

Cut the tax to zero on purpose, starting with your own visible behavior.

The playbook follows directly from what the evidence says removes the penalty. Use AI in your own work and say so specifically, because judges who use it stop punishing it, and nothing normalizes disclosure faster than the person who runs the room admitting a model wrote their first draft. Celebrate wins by method and not just outcome: when someone ships something good with AI, praise the how in public, which converts disclosure from a confession into a flex. Make "walk me through how you built this" a standard, curious question rather than an audit. And close the loop with tooling: sanction tools good enough that honesty and compliance stop being in tension. Then run one diagnostic on yourself. Count the AI-assisted wins your company celebrated openly last quarter. If the number is zero while half your workforce uses AI daily, the likeliest explanation is a tax you're collecting without noticing, and every capable person in the building has decided it's cheaper to keep paying it than to be seen paying it.

If you want an absorption strategy where your best users teach instead of hide, learn about our AI Blueprint approach or reach us at contact@theyor.com